Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Running NetBSD-current under qemu-nvmm


A couple of days ago I built wip/qemu-nvmm to find out if I can use
this setup instead of xen on one particular laptop I have in use now.
Cudos to this - my initial tries were very promising, I think I can
easily run some 8-10 VMs on this hardware. I installed a couple of
Linux distributions and FreeBSD 12, no problem here, using VNC display
I can reasonably well use this, the host does not seem under any load
when the guests are idling, so everything was OK. The network bridge
also worked first time.

The problem I am having is with NetBSD-current, amd64. I am able to
install it and run it with the default accelleration, which is
obviously very slow (perhaps 20-30 times, I haven't measured). If I
use nvmm, I consistently get a panic as follows:
 Making device nodes ...
 [ 141.4547199] fatal protection fault in supervisor mode
 [ 141.4547199] trap type 4 code 0 rip 0xffffffff8021e0ea cs 0x8
rflags 0x10246 cr2 0x7bcebdf892b0 ilevel 0x8 rsp 0xffff83802b513f98
 [ 141.4547199] curlwp 0xfffff0e0f6350240 pid 118.1 lowest kstack
 kernel: protection fault trap, code=0
 Stopped in pid 118.1 (sh) at      netbsd:xrstor+0xa      [xsave]
db{0}> bt
xrstor() at netbsd:xrstor+xa
cpu_switchto() at netbsd:cpu_switchto+0xb1

This is obviously in the last stages of the installation, the sets
have already been copied, the resulting image is not bootable at this
moment. If I carry on the installation without nvmm and then boot it
with nvmm, I get exactly the same panic when init starts.

I am new to qemu usage, so perhaps I have missed some steps in the
case of NetBSD, but with Linux it ran straight away, using a command
line like:

qemu-system-x86_64 -usbdevice tablet \
   -m 2048M -accel nvmm \
   -vnc :6 -hda /dev/rdk6 -vga vmware \
   -smp 2 -net tap,fd=4 4<>/dev/tap1 -net nic

On a separate problem I am having, not particularly important at this
stage, I installed the two Linux systems on two slices on a
GPT-labelled disk, no problem here. If I try to install NetBSD on a
similar slice, fdisk is not able to label the disk, showing some 4-5
I/O errors, the installation process is hanging and I can't even
interrupt it. I also could not install FreeBSD 12 on such a slice, but
here the attempt to label the disk failed and there was no hang.



Home | Main Index | Thread Index | Old Index